Letter Urging FDA to Review Safety and Efficacy of Pharmaceuticals Marketed Prior to 1938

May 29, 1996

David A. Kessler, M.D., J.D.
Food and Drug Administration
5600 Fishers Lane
Rockville, MD 20857

Dear Dr. Kessler:

We are writing to urge you to expeditiously begin a safety and efficacy review of all pharmaceuticals marketed in the United States prior to 1938, beginning with the mostly heavily prescribed drugs.[1] In 1978, the last year for which data are available, the FDA estimated that 240 pre-1938 pharmaceuticals were being manufactured. Of these, only 45 had submitted safety and efficacy data in New Drug Applications, in most instances not for all dosage forms of the medication. FDA's failure to conduct such a review has permitted these medications, which include Synthroid, the fourth-most commonly prescribed drug in the United States with annual wholesale sales of $276 million, to remain largely unregulated.

Because there is no FDA-approved standard for bioequivalence (suitability for substitution), Synthroid has captured 85 percent of the U.S. market for levothyroxine. Synthroid's manufacturer, Knoll Pharmaceutical Company of Mount Olive, New Jersey, is using the absence of a bioequivalence standard to protect its market share by misleading doctors and pharmacists in glossy promotional materials that proclaim that there is "No proven bioequivalent product," "There is no substitute for Synthroid" and "No adequate and well-controlled studies have demonstrated bioequivalence among levothyroxine sodium products" (see the attached advertisement from Drug Topics, April 22, 1996). These statements are only possible because the FDA has not adequately regulated levothyroxine.

FDA's failure to regulate has also permitted the outrageous chain of events related to Synthroid outlined in the Wall Street Journal on April 25, 1996. In summary, in 1986, Knoll approached researchers at the University of California, San Francisco (UCSF) to do a study comparing the bioequivalence of four brands of levothyroxine, including Knoll's own formulation, Synthroid.[2] Levothyroxine is a medication used to replace thyroid hormone in patients with low levels of the hormone (hypothyroidism). The research, completed in 1990, found the four forms of levothyroxine (two brand name and two generic) to be "bioequivalent" and a manuscript reporting this was submitted to and accepted for publication by the prestigious Journal of the American Medical Association in late 1994. However, fearing a threatened lawsuit from Knoll, which had signed a contract with UCSF preventing the researchers from publishing any data without Knoll's permission, UCSF pressured its own researchers to withdraw the manuscript one day prior to the journal's going to press. Subsequently, Knoll employees produced a new manuscript based on the UCSF data claiming that the UCSF study was flawed and that the four levothyroxine preparations were "therapeutically inequivalent." A paper reflecting this perspective was published in June 1995 under the title "Limitations of levothyroxine bioequivalence evaluation: analysis of an attempted study" in the obscure American Journal of Therapeutics, whose Associate Editor is the first author on the revised manuscript. The publication of this latter manuscript, which did not include any of the UCSF researchers as co-authors, may well preclude publication of the version accepted by the Journal of the American Medical Association.

The cost implications of this perversion of the scientific process are staggering. While most medications that lose patent protection (which usually occurs 17 years after the patent is filed) rapidly lose market share, in 1993 Synthroid still retained 85 percent or $235 million of the $276 million wholesale levothyroxine market, even though levothyroxine was initially marketed over half a century ago. If, as a result of these important findings being made public, Knoll's market share had fallen to 25 percent of prescriptions comparable to the market share of Valium about a decade after it lost patent at least $258 million (wholesale) would have been saved between 1993, by which time the results should have been published, and the present. However, thanks to Knoll's heavy-handed tactics and UCSF's failure to resist them, physicians and patients have been denied this basic information and American consumers have paid the price.

As mentioned above, another reason that Knoll has cornered such a disproportionate share of the market is that, except for standards governing Good Manufacturing Practice, the FDA has failed to regulate levothyroxine. The absence of any FDA guidelines on levothyroxine bioequivalence or on how studies to assess bioequivalence should be conducted, has permitted the perpetuation of myths of generic levothyroxine inferiority. This is not the first time levothyroxine bioequivalence has been an issue. In 1982, for example, Knoll reformulated Synthroid. Many patients ended up receiving inordinately high doses of Synthroid because the pre-1982 formulation of Synthroid had contained only about 78 percent of the expected potency.[3] Although bioequivalence standards exist for most drugs marketed in the United States, levothyroxine's status as a pre-1938 drug has allowed it to evade adequate regulation.

The events described in the Wall Street Journal date back to March 1986 when Drs. Betty J. Dong and Francis S. Greenspan at UCSF's Schools of Pharmacy and Medicine were approached by Knoll and asked to conduct a clinical trial comparing Synthroid to three other forms of levothyroxine. By February 1988, the study had been designed and had the approval of both Knoll and UCSF's Committee on Human Research. A contract to perform the work was signed in May 1988. The contract stipulated the study design and methods of statistical analysis, consistent with standard methods for bioequivalence determination. The contract also contained the following sentence: "Data obtained by the investigator while carrying out this study is [sic] also considered confidential and is not to be published or otherwise released without consent from Flint Laboratories, Inc" (later known as Knoll). The contract was so similar to other pharmaceutical company contracts being routinely approved by UCSF (and other institutions), that Dr. Dong was not required to pass it through UCSF's Contracts and Grants office prior to signing it. UCSF's legal counsel at the time, Joe Cowan, JD, reviewed the contract on several occasions after it had been signed but before problems arose with Knoll, and stated that it would not represent a barrier to publication and that the university would defend her should any problems arise. (Subsequently, in the Fall of 1995, UCSF issued guidelines urging researchers not to sign contracts with clauses precluding publication without the sponsor's approval. However, such gag clauses may still be employed at other universities.)

Between 1988 and June 1990, the study was conducted according to the guidelines described in the contract, whereupon the study blood samples were sent to the University of Chicago for analysis. These results were provided to Knoll and the UCSF authors in December 1990. It was only when the results of the study became available that Knoll began to raise questions about the conduct of the study and its interpretation, a controversy that would continue for four years. Some of Knoll's comments were incorporated into the manuscript being prepared by Dr. Dong and her colleagues. Apparently dissatisfied, on March 4, 1994 Knoll Senior Vice President for Research and Development, Neil M. Kurtz, MD, wrote to Dr. Dong and several members of the UCSF faculty and staff, including the Chancellor, all the Vice-Chancellors, and the Chairpersons of Medicine, Family and Community Medicine, and Pharmacy, as well as the Director of UCSF's Program in Medical Ethics. The letter argued that the study had been flawed and should not be published. On August 5, 1994, UCSF Associate Vice-Chancellor for Academic Affairs Karl J. Hittleman, PhD, wrote to Knoll's Dr. Kurtz refusing to suppress the manuscript and saying that to do so would be a violation of academic freedom. In November 1994, the manuscript was accepted by the Journal of the American Medical Association.

Dr. Dong's manuscript, entitled "Bioequivalence of generic and brand levothyroxine products in the treatment of hypothyroidism" was a randomized, four-way crossover trial comparing the bioavailability of four levothyroxine preparations: Synthroid (Knoll), Levoxine (Daniels), and two generic formulations manufactured by Pharmaceutical Basics and distributed by Geneva Generics and Rugby. Subjects received all four brands of levothyroxine for a period of at least six weeks in a randomly assigned sequence that ensured that potential carryover effects from the previous formulation would introduce no bias into the evaluation. Primary investigators were blinded as to which formulation the subjects were taking at any given time.

Three aspects of bioequivalence (area under the curve, peak serum concentration, and time to peak serum concentration) were measured for each of three indices of thyroid function (T4 (levothyroxine), T3, and free T4 index). For all nine comparisons, there was no statistically significant difference between the four formulations, indicating that they were bioequivalent, even using a criterion for bioequivalence considerably more stringent than that employed by the FDA for other compounds. The authors concluded that the four generic and brand levothyroxine preparations studied were bioequivalent and interchangeable for most patients taking levothyroxine hormone.

In the Fall of 1994, Gilbert H. Mayor, MD, Senior Director of Medical Research at Knoll, called Dr. Greenspan and threatened a lawsuit against both UCSF and the authors should the manuscript be published. In December 1994, Dr. Dong's paper was accepted for publication in the Journal of the American Medical Association following review by five peer reviewers, more than usually review manuscripts for the Journal. One day prior to the date on which that issue of the journal was to go to print, and with her article already in galleys, Dr. Dong received a telephone call from Cynthia Lynch, JD, assistant to Shelley Drake, the campus counsel who had succeeded Joe Cowan in dealing with this issue. Ms. Lynch insisted that Dr. Dong withdraw the article from publication, and stated that UCSF would not represent the authors should they be sued by Knoll. Consequently, Dr. Dong was compelled to contact the journal and ask that the article be removed.

The Journal has stated verbally that it would accept the article if it were resubmitted. However, the threat of a lawsuit and UCSF's refusal to defend its own researchers against such a suit have dissuaded the researchers from resubmitting. Dr. Dong has sought redress within the UCSF system, but in the Fall of 1995 was denied a hearing on this issue before the university's Promotion and Tenure Committee.

In June 1995, the American Journal of Therapeutics, a journal not even listed in the medical journal computer database MedLine, published an article entitled "Limitations of levothyroxine bioequivalence evaluation: analysis of an attempted study," which was based upon the data generated by Dr. Dong and her colleagues. The authors were two Knoll employees, Drs. Mayor and Kurtz, and Dr. Tony Orlando of Pharmaco LSR in Austin, TX, who had previously worked for Knoll. Dr. Mayor is one of two Associate Editors of the American Journal of Therapeutics and also has responsibility for its endocrinology section. None of the UCSF researchers were included as authors, nor were they included in a lengthy Acknowledgments section.

The article came to markedly different conclusions from Dr. Dong's paper. Using the same data as those used by the UCSF researchers, the Knoll researchers assessed thyroid function using two methods: 1) with no correction for baseline values (the method used by the UCSF researchers); and 2) with correction for baseline values.[4] The first analysis came to similar conclusions to Dr. Dong: the formulations "[appeared] to be bioequivalent." The second analysis, however, resulted in many findings of differences between the four products, suggesting lack of bioequivalence. The article went on to list in exhaustive detail various purported limitations of the study and, in a highly unusual half-page table, listed these "major study limitations." The study, the article said, had "encountered numerous and often unpredictable difficulties in execution" and these difficulties "were too numerous and substantive to consider this a definitive bioequivalence study." The authors concluded that "the levothyroxine products studied, although appearing to be bioequivalent, were, in fact, therapeutically inequivalent."

There are two major reasons for the apparent discrepancy between the two papers: 1) the issue of baseline correction; and 2) the Knoll paper also included data on another index of thyroid function, TSH, the measurement of which produced most of the differences between the formulations.[5] The net effect of these statistical manipulations was to call the UCSF paper's findings into question. This would not have been possible had the FDA exercised its authority to adequately regulate levothyroxine, including establishing clear standards for bioequivalence determination.

In sum, Knoll sponsored a research study and, when it found the conclusions to not be to its taste, embarked upon a campaign of scientific distortion and legal threats. Six years after the study's completion, and over one year since it passed peer review in a prestigious medical journal, the investigators' manuscript has still not been published, and, due to Knoll's publication of its own self-serving article, may now not be publishable. These events would not have occurred had the FDA adequately regulated levothyroxine and provided adequate guidelines on bioequivalence determination. The FDA has essentially permitted the pharmaceutical industry to regulate itself in this area, with predictable consequences. However, the suppression of these findings, which have enormous implications for the practice of clinical medicine and cost-containment, would not have occurred had UCSF not failed to review the contract and then succumbed to the gag clause, thereby sounding a retreat from the time-honored principle of academic freedom. The public has been ill-served by both of these institutions. Now it is time for the FDA to prevent any further pharmaceutical company manipulation by extending the safety and efficacy regulations to include drugs marketed prior to 1938. Hundreds of millions of consumer dollars a year are at stake.


Peter Lurie, M.D., M.P.H.
Research Associate
Public Citizen's Health Research Group

Sidney M. Wolfe, M.D.
Public Citizen's Health Research Group


1. In 1962, amendments to the Food, Drug, and Cosmetic Act required that safety and efficacy data be provided for all pharmaceuticals, with the exception of those marketed prior to 1938. However, if the manufacturer wished to change the indications or dosage form of a pre-1938 drug, a full safety and efficacy application, known as a New Drug Application, would have to be submitted.

2. During the course of the events described here, the company was known first as Flint Laboratories, Inc., then as Boots Pharmaceuticals, Inc. and finally as Knoll Pharmaceutical Company. For simplicity, these are all referred to as "Knoll" in this letter.

3. Stoffer SS, Szpunar WE. Potency of levothyroxine products. Journal of the American Medical Association 1984; 251:635-636.

4. Correcting for baseline, although controversial, is not customary in assessing the bioequivalence of drugs that are natural substances produced by the human body. In diseases such as hypothyroidism, such substances may still be produced by the body, but at lower than normal levels. Indeed, thyroid hormone production by the thyroid gland is essentially turned off when replacement thyroid hormone is given. Due to the impossibility of distinguishing the body's own hormone from that in the drug on chemical assays, baseline correction can not easily be done. In addition, the complex feedback mechanisms that the body uses to regulate hormone levels make the uncorrected measurement the one most consistent with how the body actually functions. One result of inappropriate baseline correction can be negative values for the corrected variables, as occurred in the Knoll paper, a finding that has no biological meaning.

5. TSH would probably not be included in any FDA requirement for establishing the bioequivalence of levothyroxine. This is because establishing bioequivalence does not require demonstrating that the drug is effective (this should already have been done by the brand name company); the generic company need only demonstrate that similar quantities of the drug are released into the blood stream. Thus T4, T3, and free T4 index would likely be components of levothyroxine bioequivalence testing, but TSH, which measures levothyroxine effectiveness, would not.